Parsemux

Document parser orchestrator — auto-routes to the optimal OSS parser for each document.

Try the demo | API Docs | GitHub

Why Parsemux?

• Auto-routing — detects document type and picks the best parser automatically
• 5 parsers, 1 interface — PyMuPDF, Kreuzberg, Docling, MinerU, Marker
• Every interface — CLI, REST API, MCP server, Web UI
• Image extraction + VLM description — BYOK, auto-detects provider from key prefix
• Compare mode — run all parsers on the same doc, pick the best output
• Cost comparison — shows savings vs AWS Textract, Google Document AI, etc.
• Zero config — pip install parsemux[pymupdf,cli] and go

Install

curl -fsSL https://raw.githubusercontent.com/vericontext/parsemux/main/install.sh | sh

Or with pip:

pip install parsemux[pymupdf,kreuzberg,cli]

Usage

Parse a document

parsemux parse document.pdf                          # auto-route, markdown output
parsemux parse document.pdf --parser kreuzberg       # explicit parser
parsemux parse document.pdf --format json            # JSON output
parsemux parse document.pdf --extract-images         # extract images as base64
parsemux parse document.pdf --dry-run                # preview routing without parsing
parsemux parse ./docs/ --batch                       # batch directory

Image description with VLM (BYOK)

parsemux parse doc.pdf --extract-images --describe-images --vlm-key sk-...

Provider is auto-detected from key prefix (sk- → OpenAI, sk-ant- → Anthropic, AI → Google). Default models: gpt-5.4-nano, claude-haiku-4.5, gemini-2.5-flash, qwen2.5vl:7b (local).

Start your own server

parsemux serve                    # REST API at :8000 + MCP at /mcp

MCP server

# Local (Claude Desktop, Cursor — stdio transport)
parsemux mcp

# Remote (Streamable HTTP)
parsemux mcp --remote --port 8000

Claude Desktop config (local):

{
  "mcpServers": {
    "parsemux": {
      "command": "parsemux",
      "args": ["mcp"]
    }
  }
}

For AI agents

parsemux schema                           # machine-readable command schemas
parsemux parse doc.pdf --dry-run          # preview routing
parsemux list-parsers --json              # available parsers as JSON
parsemux detect doc.pdf --json            # MIME + recommended parser

Supported Parsers

Parser	Best For	Speed
PyMuPDF	Digital PDFs	1,000+ pages/sec
Kreuzberg	91+ formats, OCR	Rust core
Docling	Tables (97.9%)	CPU
MinerU	Scanned docs	GPU
Marker	Batch + LLM-enhanced	GPU

Cloud Demo vs Self-hosted

	Self-hosted	Demo
File limit	100 MB	10 MB
Rate limit	None	10 req/min
MCP remote	/mcp	Disabled
VLM key	.env	BYOK (enter in UI)

Docker

docker compose up
# API: http://localhost:8000
# MCP: http://localhost:8000/mcp

Contributing

Contributions welcome!

1. Fork the repo
2. Create a feature branch (git checkout -b feat/my-feature)
3. Run tests (pytest -q)
4. Open a PR

See issues for ideas.

Acknowledgments

Built on the shoulders of these excellent open-source parsers:

• PyMuPDF — blazing-fast digital PDF parsing
• Kreuzberg — Rust-powered multi-format + OCR
• Docling — best-in-class table extraction
• MinerU — high-quality scanned doc pipeline
• Marker — LLM-enhanced batch conversion

License

MIT